Consier speeding up get_secure_random_bytes

[TheBlueMatt]

We really should swap out for a "real" CSPRNG, not because it'll be more secure, but because it'll be way faster than deriving a new HD key and hashing it. Preferrably something chacha-based given we already have chacha in-tree.

[danielgranhao]

I can take this up if no one has started yet.

I've done an initial experiment, and it seems that using chacha can get the execution time of get_secure_random_bytes from ~235 ns down to just ~4 ns on my machine.

Some of the tests broke, but I think it's due to them relying on the values that the original method generated using the same starting_time_nanos and starting_time_secs. Looking into it.

[TheBlueMatt]

Nice! Thanks.

[danielgranhao]

I think I need some help here. I opened a draft PR with the current state of my attempt -> #1974.

One unit test fails because 2 Outpoints are equal when they should be different. I'm not sure if this shows a problem with the get_secure_random_bytes() method or if it's something that must be changed in the tests. I've already spent some time trying to figure it out but to no avail...

Also, I'm not sure of the safety of the RNG implementation. Some thoughts on that would also be appreciated 🙂

[TheBlueMatt]

Let's continue the discussion on the PR, thanks!